The standard cosmological model based on cold dark matter (CDM) predicts a large number of subhalos for each galaxy-size halo. It is well known that matching the subhalos to the observed properties of luminous satellites of galaxies in the local universe poses a significant challenge to our understanding of the astrophysics of galaxy formation. We show that the cosmic evolution and host mass dependence of the luminosity function of satellites provides a powerful new diagnostic to disentangle astrophysical effects from variations in the underlying dark matter mass function. We illustrate this by comparing the results of recent observations of satellites out to $z=0.8$ based on Hubble Space Telescope images with the predictions of three different sets of state-of-the art semi-analytic models with underlying CDM power spectra and one semi-analytic model with an underlying Warm Dark Matter (WDM) power spectrum. We find that even though CDM models provide a reasonable fit to the local luminosity function of satellites around galaxies comparable or slightly larger than the Milky Way, they do not reproduce the data as well for different redshift and host galaxy stellar mass. This tension indicates that further improvements are likely to be needed in the description of star formation if the models are to be reconciled with the data. The WDM model matches the observed mass dependence and redshift evolution of satellite galaxies more closely than any of the CDM models, indicating that a modification of the underlying power spectrum may offer an alternative solution to this tension. We conclude by presenting predictions for the color magnitude relation of satellite galaxies to demonstrate how future observations will be able to further distinguish between these models and help constrain baryonic and non-baryonic physics.